Objects with wide range of contrasts makes thresholding very difficult

Hello everybody,

im currently working on my Master Thesis, developing an automated counting approach for Collembola from pitfalltraps with Fiji.

As you can seefrom the provided image, some specimens are almost translucent, while others have good contrast. Is there any way that i can separate all individuals (via thresholding or other means) no matter their level of contrast from the background so I could count them later?
Your help will be much appreciated!
Cheers
Alex

If you are using imagej you can try Image->Adjust->Auto Local Threshold

Make sure you set it to look for dark objects and also set an appropriate radius, you may have to play around with that parameter a bit to find the best value for your image.

If you use Python there is good documentation on the thresholding functions here including an example that shows how to apply a local threshold.

Thank you very much, that actually seems to work quite well!!!

Cheers!

Alex

1 Like

A little follow up:

This is what I was able to produce:


great progress!

Cheers!

Alex